This disclosure addresses the single-image compressive sensing (CS) and reconstruction problem. A scalable Laplacian pyramid reconstructive adversarial network (LAPRAN) facilitates high-fidelity, flexible and fast CS image reconstruction. LAPRAN progressively reconstructs an image following the concept of the Laplacian pyramid through multiple stages of reconstructive adversarial networks (RANs). At each pyramid level, CS measurements are fused with a contextual latent vector to generate a high-frequency image residual. Consequently, LAPRAN can produce hierarchies of reconstructed images and each with an incremental resolution and improved quality. The scalable pyramid structure of LAPRAN enables high-fidelity CS reconstruction with a flexible resolution that is adaptive to a wide range of compression ratios (CRs), which is infeasible with existing methods.
Legal claims defining the scope of protection, as filed with the USPTO.
1. A method for compressively encoding a source image, the method comprising: randomly encoding the source image into multi-rate measurements, from which compressed image data is obtained, the randomly encoding comprising: producing first compressed image data at a first compression ratio; and producing second compressed image data at a second compression ratio; and providing the compressed image data to a system configured to reconstruct the source from the compressed image data by performing actions comprising: inputting the first compressed image data to a first reconstructive adversarial network (RAN), from which a first reconstruction of the source image at a first resolution is obtained; and inputting the first reconstruction and the second compressed image data to a second RAN, from which a second reconstruction of the source image at a second resolution is obtained.
2. The method of claim 1, wherein the first compressed image data is produced using a first sensing matrix ϕ1.
3. The method of claim 2, wherein the second compressed image data is produced using a second sensing matrix ϕ2.
4. The method of claim 3, wherein the second sensing matrix ϕ2 includes the first sensing matrix ϕ1.
5. The method of claim 1, wherein randomly encoding the source image into the multi-rate measurements further comprises, for each of k image reconstruction stages, wherein k is an integer, producing i-th compressed image data at an i-th compression ratio, wherein i is an integer.
6. The method of claim 5, wherein each i-th compressed image data comprises a smaller number of compressive sensing measurements than each (i+1)-th compressed image data.
7. The method of claim 6, wherein each i-th compressed image data comprises at least ¼ the number of compressive sensing measurements of each (i+1)-th compressed image data.
8. The method of claim 6, wherein each i-th compressed image data comprises ½ the number of compressive sensing measurements of each (i+1)-th compressed image data.
Cooperative Patent Classification codes for this invention. Click any code to explore related patents in that topic.
October 11, 2022
January 28, 2025
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.